대한조선학회논문집 Vol. 48, No. 6, pp.

552-559, December 2011

Journal of the Society of Naval Architects of Korea http://dx.doi.org/10.3744/SNAK.2011.48.6.552

자유낙하식 구명정의 가속도 응답 추정을 위한

LS-DYNA 에서의 다중물질 ALE 와 단일물질
ALE의 비교
배동명1,†․ 자키1
1
부경대학교 조선해양시스템공학과

Comparisons of Multi Material ALE and Single Material ALE in LS-DYNA

for Estimation of Acceleration Response of Free-fall Lifeboat
Dong Myung Bae1,† ․ Ahmad Fauzan Zakki1
1
Department of Naval Architecture and Marine Systems Engineering, Pukyong National University

Abstract
An interest in Arbitrary Lagrangian Eulerian (ALE) finite element methods has been increased due to more accurate responses in
Fluid-Structure Interaction(FSI) problems. The multi-material ALE approach was applied to the prediction of the acceleration response of
free-fall lifeboat, and its responses were compared to those of the single-material ALE one. It could be found that even though there
was no big difference in the simulation responses of two methods, the single-material and multi-material ALE ones, the latter
multi-material ALE method showed a little bit more close response to those of experimental results compared to the former
single-material ALE one, especially in the x- and z-direction acceleration responses. Through this study, it could be found that several
parameters in the ALE algorithms have to be examined more carefully for a good structural safety assessment of FSI problems.

Keywords : Free-fall lifeboat (자유낙하식 구명정), Acceleration response (가속도 응답), Multi material ALE (다중물질 ALE), Single material
ALE (단일물질 ALE)

1. Introduction necessary for the continuous calculation. Since the fluid part
is extremely distorted during the launch of free-fall lifeboat in
the water, the Lagrangian method is not suitable for the fluid
During the launch of free-fall lifeboat, there are two model.
primary concerns. The first one is the motion of the lifeboat Since the material part is treated differently from the
which is affected by the change in hull shape, weight geometric mesh in Arbitrary Lagrangian Eulerian (ALE) finite
distribution and initial conditions. element method contrary to the Lagrangian one, while the
The second one is the acceleration responses to which lifeboat structure hit the water surface, the fluid element will
the occupant in the lifeboat is subjected. Harmful be deformed largely. This phenomenon will make the time step
accelerations may occur in the free-fall lifeboat when it changed to a very small value for explicit calculations. The
impacts on the water. The hydrodynamic impact of boat ALE methods or rezoning are used to create a new
during water entry is a complex problem and makes the undistorted mesh for the fluid domain, which allows the
establishment of the numerical analysis be a challenging task. calculations to continue.
Lagrangian method is usually used for the structural The ALE method has been applied to many
analysis and occasionally for the fluid one. Its computational Fluid-Structure Interaction (FSI) problems, such as slamming
cost could depend on the mesh size and deformation. If large (Lee, et al., 2008a,b, 2010) and sloshing (Lee, et al., 2010),
deformations occurred, the mesh would be high distorted and and it is recommended that their simulation responses should
the calculation could be terminated. Remesh process is be verified by the experimental ones. These FSI problems

접수일 : 2011년 2월 11일 󰠛 1차 수정일 : 2011년 10월 13일 󰠛 게재확정일 : 2011년 11월 7일
†
교신저자 : dmbae@pknu.ac.kr, 051-629-6613
 배동명․ 자키

could be conveniently simulated using Arbitrary Lagrangian nodes to converge to an uniform mesh by solving
Eulerian (ALE) formulation and Euler–Lagrange coupling equipotential equation for nodal displacement.
algorithm of LS-DYNA code (LSTC, 2009). Volume of Fluid There are two approaches to implement the ALE equations,
(VOF) method is adopted for solving a broad range of such as the solution of the fully coupled equations for
nonlinear free surface problems and coupling algorithm is computational fluid mechanics handling a single material in an
more suitable for the FSI problems with very complicated element and reference of an operator split for each time step
structure, where fluid grid can overlap the structural mesh into two phases with the first Lagrangian phase and the
(Aquelet, et al., 2003, 2006; Souli, et al., 2000). second advection phase. In the advection phase, transport of
The single material ALE approach was applied to the mass, internal energy and momentum across cell boundaries
prediction of the acceleration response of free-fall lifeboat are computed. This might be thought of as remapping the
(Bae et al, 2010). In this paper, the Multi Material ALE displaced mesh at the Lagrangian phase back to its original
or arbitrary position element. The VOF method is attractive
method was adopted for the same FSI problem and its
because of being applicable to solve broadly non-linear
responses were compared to those of the single-material ALE
problems in fluid and solid mechanics. The method allows
one.
arbitrary large deformations and enables free surfaces to
evolve (Aquelet, et al., 2003, 2006; Souli, et al., 2000).

2. Arbitrary Lagrangian Eulerian

(ALE) Method in LS-DYNA 3. Multi Material ALE

The ALE method in LS-DYNA code is the computational The multi-material ALE formulation uses the same
algorithm that applies the conservation equations in the finite governing equations as the single-material ALE formulation
element method. Energy, mass, and momentum are conserved except that the calculations are performed iteratively for each
and advected from element to element. LS-DYNA uses a split material group in the element. For material group i, the strain
operator technique to solve the conservation equations using rate is  , and the internal energy is Ei. The deviatoric or
Eulerian formulation and ALE one. The conservation of mass, shear stress is  , and pressure from the equation of state is
momentum and energy are given by the following equations:  . Then the stress is given by the following equation (4).

         (4)
    ÷           (1)
  
    The internal nodal force is calculated by the equation (5).
              (2)
  
 
             
  
(3)
     

  (5)

where,   is the strain-displacement matrix.  is the

where,  is the material density,  is the material velocity,
volume fraction in the element and has the following
 is stress tensor,  is the strain tensor,  is internal energy
requirement;
and  is the mesh velocity.
While the fluid element has extremely distorted during the
 
analysis, mesh rezoning algorithms are used to maintain the

 
   (6)
fluid mesh, and computational time does not increase. While a
non uniform mesh is transformed into an uniform mesh in the
rezoning method, a non uniform mesh, reversely into an where, Ngroups is the number of different materials in the
uniform mesh in the equipotential algorithm. The latter element. The volume fractions need to be recomputed using a
equipotential algorithm maintains the location of the boundary pressure relaxation in each element. The pressure relaxation is
nodes at their Lagrangian location and iterates the internal given as follows;

JSNAK, Vol. 48, No. 6, December 2011 553

자유낙하식 구명정의 가속도 응답 추정을 위한 LS-DYNA 에서의 다중물질 ALE 와 단일물질 ALE의 비교

    Table 1 Principal dimension of free-fall lifeboat

           (7)
  Length 7950 mm
       (8) Breadth 3070 mm
Draught 1550 mm
where,  is the amount of material (volume) flux in the Weight 5112 kg
element after the Lagrangian calculation and the subscripts Max Occupant 35 person
refer to adjacent elements. The amount of total material to be LCG 50% From AP
advected can now be calculated. The volume of fraction is
used to determine whether the way of materials in the element
are advected or not. The rules of advection in LS-DYNA as
follow;

Rule 1:
Material common to both neighboring elements is
transported in proportion to the volume of the acceptor
element, with the restriction that no more than the total donor
element volume may be transported.

Rule 2:
If there is no remaining material common to both elements,
Fig. 1 Configuration of F.E. model of free-fall lifeboat
the remaining transport volume is proportional to the remaining
with Multi Material ALE
donor volume.
The process is repeated for each material at every time
There are several commands and options for the fluid
step. As the number of materials to be tracked in each
modeling and coupling algorithm using FSI analysis technique
element increases, the computational expense will increase.
of LS-DYNA code in addition to the element is usually
considered with ELFORM 11 for the multi-material ALE in
SECTION_SOLID command, contrary to the case of
4. Multi Material ALE Modeling single-material ALE with ELFORM 12.
For the fluid material description, MAT_NULL command and
The free-fall lifeboat launching were simulated by ALE3D Equation of State (EOS) have to be defined (LSTC, 2009).
option of LS-DYNA code using single-material ALE approach. Since this study is not concerned with tracking the
The outer surface of the lifeboat modeling was modeled using propagation of energy and pressure in water and air. EOS
4,805 rigid quadrilateral shell elements. Since the main linear polynomial was used for the properties of water and
objective of this study is to predict the acceleration response, air, the property of EOS linear polynomial of fluid model is
the rigid lifeboat model could be acceptable. The penalty shown in Table 2.
method was adopted for solving the contact problem between Several parameters are very sensitive to the coupling between
the lifeboat and skid. the fluid and structure in CONSTRAINED_LAGRANGE_IN_SOLID
The fluid model consists of two parts, such as air and command. Coupling leakage and penalty forces, etc., are
water, using 674500 hexagonal Eulerian elements, whose affected by the penalty factor, the number of quadrature
dimensions are 42.7 × 15.1 × 32.0m for the air and 42.7 coupling points on a Lagrangian segment and the mesh size
× 15.1 × 13.0m for the water domain, respectively, as
ratio of the structure mesh to fluid. Following this conditions,
shown in Fig. 1. The mesh size of the fluid are 0.3 × 0.3
the default values are used for the penalty factor and the
× 0.3m around at the surface, and is increased
number of quadrature coupling points. Additionally, continuum
proportionally to vertical upward and downward directions
treatment and advection method can be selected in
with 20% bias. The principal dimension of lifeboat is
CONTROL_ALE command.
summarized in Table 1.

554 대한조선학회논문집 제 48 권 제 6 호 2011년 12월

 배동명․ 자키

Table 2 EOS linear polynomial of fluid model

Item Water Air
density (kg/m3) 1025 1.225
C0 (Pa) 0 0
C1 (Pa) 2.002e9 0
C2 (Pa) 8.436e9 0
C3 (Pa) 8.010e9 0
C4 0.4394 0.4
C5 1.3937 0.4
C6 0 0
E0 (Pa) 2.086e5 2.5e5
V0 1 1 Fig. 2 Acceleration response in x-axis direction according
to fluid mesh size

The boundary condition of fluid model and constraint

condition of structure are also important to the acceleration
6. Comparison Multi Material
responses of free-fall lifeboat water entry on to the water. ALE to Single ALE
The following assumptions are considered as follows:
The numerical simulations were carried out using LS-DYNA
1. Only gravitational external load is applied to the whole Version 9.71 R4.2 with single precision. Based on the
system using a load curve for the gravitational computational time, the case of multi material ALE took longer
acceleration time history. than that of single material ALE, as shown in Table 3. This
2. Top, side and bottom boundaries of the fluid are fixed can be explained that multi material ALE needs more
to the normal directions. computational procedure, since the computation is regarding
3. Initial velocity of lifeboat is set to zero. to the interaction of a lifeboat with the water and the air.

5. Sensitivity of FE Model in FSI Table 3 Computational time in each case

Analysis Method Computational time

Multi Material ALE 8 hrs 41 mins

In this FSI analysis, lifeboat was modeled by Lagrangian
shell elements, and fluid, by Eulerian solid elements. In Single Material ALE 2 hrs 49 mins

general, it is known that the coupling algorithm is usually

sensitive to the relative mesh size between Lagrangian and The acceleration response of lifeboat was used for the
Eulerian elements. For the investigation of the sensitivity of assessment of the occupant injury potential. Since the
their relative mesh size, three mesh sizes were considered, experiment data was taken at the data record frequency 2500
such as 0.2 mm (model I), 0.3 mm (model II) and 0.5 mm Hz and filtered with Butterworth low pass filter 20 Hz,
(model III), where mesh size of lifeboat was 0.1 mm. simulation results with frequency scale 1000 Hz were filtered
Fig. 2 shows the acceleration response in X axis direction using a Butterworth digital 8.0 Hz low pass filter. The
according to fluid mesh size. It could be found that the acceleration responses of simulation for the x-axis, y-axis
response of model II was the most close to that of and z-axis directions are compared with those of experiment,
experiment, as shown in Fig. 2. Much leakage occurred in as shown in Figs. 3∼6, according to loading conditions, such
models II and III compared to model I. From this study, the as full, 50% forward, 50% backward and empty loading
relative mesh size between the lifeboat Lagrangian and fluid conditions, respectively. Their peak values are also
Eulerian elements should be determined for the reasonable summarized in Table 4~6.
response with less leakage. Based on the experimental results, it was found that the

JSNAK, Vol. 48, No. 6, December 2011 555

자유낙하식 구명정의 가속도 응답 추정을 위한 LS-DYNA 에서의 다중물질 ALE 와 단일물질 ALE의 비교

most severe responses were obtained from 50% backward little bit more close to those of experimental ones compared
loading condition, which could be explained that the CG of to the single material ALE case. The first large peak response
loading condition was shifted to the backward. While the shift might be caused by the bow impact to the water at the
of CG to 50% backward caused the righting moment arms to vicinity 3.0 sec.
be increased and also led to increase the severity of
slamming phase, the situations were conversed by its shift to Table 5 Maximum y-axis acceleration
50% forward.
y-axis acceleration in G
Regarding the characteristics of the longitudinal (x-axis Loading
Condition Single
direction) acceleration responses, the multi material ALE case EXP Multi ALE
ALE
in 50% backward loading condition was better than the single
material ALE case, as shown in Figs. 3~6(a) and Table 4. Full (100%) 0.871 0.411 0.394

From the experimental responses, it can be found the bow 50% Forward 0.525 0.442 0.436
entry into the water at the vicinity of 3.0 sec and then its 50% Backward 0.295 0.397 0.531
rebounce from the water at 0.4 sec later.
Empty 0.402 0.603 0.552

Table 4 Maximum x-axis acceleration

Their peak values and the interval time of the second ones
x-axis acceleration in G
Loading might depend on their CG positions according to the loading
Condition Single
EXP Multi ALE conditions, and their entry depths into the water due to their
ALE
weights. Although the second peak responses were not be
Full (100%) 2.671 3.415 3.221
found in the both method of simulation results, the simulation
50% Forward 2.752 3.642 3.566 results are still reliable as the reference of lifeboat
50% Backward 2.041 3.246 2.239 performance, because their peak values are still below the
IMO standards in X-Axis acceleration below 15g, and the
Empty 3.481 4.929 4.516
other direction one below 7g (IMO, 2003).

However, this rebounding phenomenon could not still be

Table 6 Maximum z-axis acceleration
found in the multi material ALE analysis. Even though the
peak values of both methods shows relatively larger than z-axis acceleration in G
Loading
those of experiment results, these values are still acceptable Multi
Condition EXP Single ALE
because of much lower peak values in X-direction compared ALE

to IMO standards, 15g (IMO, 2003). Full (100%) 3.963 4.641 3.555
As can be seen in Figs. 3∼6(b) and Table 5, the 50% Forward 3.340 2.971 3.482
transverse (y-axis direction) acceleration responses were 50% Backward 4.502 4.592 4.536
shown to be very small in the experimental results, while just a
Empty 3.778 3.031 3.829
small track of bow impact was shown at the vicinity of 3.0
sec in the case of full loading condition, as shown in Fig.
3(b). Even though there was no big difference in the simulation
Relatively very small tracks of bow impact also appeared at responses of two methods, the single-material and
the vicinity of 3.0 sec in the all loading conditions of both multi-material ALE ones, the latter multi-material ALE method
simulation results. It could be found that there was no showed a little bit more close peak values to those of
significant difference between simulation results of two experimental results compared to the former single-material
methods in the y-axis direction acceleration responses due to ALE one, especially in the x- and z-direction acceleration
very small responses. responses. CAR Indexes in all loading conditions of
It could be also found, as shown in Figs. 3∼6(c) and experimental and simulation results were also found to be
Table 6, that the characteristics of vertical (z-axis direction) accepted by the IMO standards with less than 1.0 (IMO,
acceleration responses of the multi material ALE could show a 2003), as shown in table 7.

556 대한조선학회논문집 제 48 권 제 6 호 2011년 12월

 배동명․ 자키

(a) x-axis direction (a) x-axis direction

(b) y-axis direction (b) y-axis direction

(c) z-axis direction (c) z-axis direction

Fig. 3 Acceleration response in full loading condition Fig. 4 Acceleration response in 50% forward loading condition

JSNAK, Vol. 48, No. 6, December 2011 557

자유낙하식 구명정의 가속도 응답 추정을 위한 LS-DYNA 에서의 다중물질 ALE 와 단일물질 ALE의 비교

(a) x-axis direction (a) x-axis direction

(b) y-axis direction (b) y-axis direction

(c) z-axis direction (c) z-axis direction

Fig. 5 Acceleration response in 50% backward loading condition Fig. 6 Acceleration response in empty loading condition

558 대한조선학회논문집 제 48 권 제 6 호 2011년 12월

 배동명․ 자키

Table 7 Combined acceleration ratio index Bae, D. M, et al, 2010. Estimation of acceleration reponse of
CAR Index free-fall lifeboat using FSI analysis technique of LS-DYNA
Loading
Single Code. Journal of the Society of Naval Architects of Korea,
Condition EXP Multi ALE
ALE 47(5), pp.681-688.
Full (100%) 0.606 0.703 0.554 IMO, 2003, International life-saving appliances code. IMO.
50% Forward 0.517 0.463 0.555 Lee, S.G. Lee, I.H. & Baek, Y.H., 2010. Wet Drop Impact
50% Backward 0.659 0.693 0.667
Response Analysis of Cargo Containment System in
Membrane-type LNG Carrier using FSI Technique of
Empty 0.590 0.550 0.629
LS-DYNA. Proceedings of the 20th International Offshore
and Polar Engineering Conference, Beijing, China, 3,
7. Conclusions pp.206-214.
Lee, S.G. Hwang, J.O. & Kim, W.S., 2008a. Wet Drop
The multi material ALE approach was applied to the Impact Response Analysis of CCS in Membrane Type LNG
prediction of the acceleration responses of 1/5 scaled 35 Carriers I: Development of Numerical Simulation Analysis
persons free-fall lifeboat model using FSI analysis technique Technique through Validation. Journal of the Society of
of LS-DYNA code, and its responses were also compared to Naval Architects of Korea, 45(6), pp.726-734.
those of the single-material ALE one. It was found that even Lee, S.G., Hwang, J.O., & Kim, W.S., 2008b, Wet Drop
though there was no big difference in the simulation Impact Response Analysis of CCS in Membrane Type LNG
responses of two methods, the single-material and Carriers II: Consideration of Effects on Impact Response
multi-material ALE ones, the latter multi-material ALE method Behaviors. Journal of the Society of Naval Architects of
showed a little bit more close peak values to those of Korea, 45(6), pp.735-749.
experimental results compared to the former single-material Lee, S.G. Lee, I.H. & Hong Anh, 2010. Numerical Simulation
ALE one, especially in the x- and z-direction acceleration of Sloshing using LS-DYNA code. Proceedings of the
responses. Annual Autumn Meeting, SNAK, Changwon, pp. 463-477.
Through this study, it could be found that several LSTC, 2009. LS-DYNA User's Manual, Version 971 R4.
parameters in the ALE algorithms have to be examined more livermore Soft Technology Corp., USA.
carefully for a good structural safety assessment of FSI Souli, M. Ouahsine, A. & Lewin, L., 2000. ALE formulation for
problems. fluid-structure interaction problems. Computer Methods in
Applied Mechanics and Engineering, 190(5), pp.659-675.
References

Aquelet, N. Souli, M. Gabrys, J. & Olovsson, L., 2003. A new

ALE formulation for sloshing analysis. Structural Engineering
and Mechanics, 16(4), pp.423-440.
Aquelet, N. Souli, M. & Olovsson, L., 2006. Euler–Lagrange
coupling with damping effects: Application to slamming
problems. Computer Methods in Applied Mechanics and
Engineering, 195(1-3), pp.110-132. 배동명 A.F.Zakki

JSNAK, Vol. 48, No. 6, December 2011 559